Digitala Vetenskapliga Arkivet

Change search
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf
Influence of Graph Structures on Performances in Graph Neural Network for Supervised Anomaly-based Network Intrusion Detection for Botnet Attacks in IoT Environments
KTH, School of Electrical Engineering and Computer Science (EECS).
2024 (English)Independent thesis Advanced level (degree of Master (Two Years)), 20 credits / 30 HE creditsStudent thesis
Abstract [en]

With the rise of Internet of Things and the development of more complex and connected networks, cyber threats have grown into a major challenge in private networks for confidentiality and integrity of data. Intrusion Detection Systems are devices or software analysing flows of data to detect malicious activities, nowadays important amounts of data are exchanged and malicious flows need to be detected. Graph Neural Networks have demonstrate promising results for Intrusion Detection as they embed structural pattern detection compared to rule- based Intrusion Detection Systems or other Machine Learning models. This thesis investigate the potential of various graph properties in Graph Neural Networks implementation in the context of high imbalance dataset for botnet attacks in Internet of Things type of environment. Featureless models rely on the strong network structural patterns of botnet attacks, while heterogeneous models intend to capture more complex patterns by differentiating several types of nodes and edges in graphs. The potential of Temporal Graph Neural Network is also investigated on dynamic graphs. The comparison of results of the studied properties allows to discuss the potential of Graph Neural Network implementations for Intrusion Detection in the context of high imbalanced datasets. As Internet of Things networks become more complex, traditional Intrusion Detection Systems struggle to detect evolving cyber threats, especially in highly imbalanced data environments. This thesis investigates the potential of various graph structures of Graph Neural Network models to enhance performances for detection of botnet attacks, offering both theoretical advancements and practical solutions for securing IoT networks.

Abstract [sv]

I och med utvecklingen av Internet of Things och mer komplexa och uppkopplade nätverk har cyberhot blivit en stor utmaning i privata nätverk när det gäller konfidentialitet och dataintegritet. Intrångsdetekteringssystem är enheter eller programvara som analyserar dataflöden för att upptäcka skadliga aktiviteter. I dag utbyts stora mängder data och skadliga flöden måste upptäckas. Grafneurala nätverk har visat lovande resultat för intrångsdetektering eftersom de innehåller strukturell mönsterdetektering jämfört med regelbaserade intrångsdetekteringssystem eller andra maskininlärningsmodeller. Denna avhandling undersöker potentialen hos olika grafegenskaper i implementeringen av grafneurala nätverk i samband med dataset med hög obalans för botnetattacker i Internet of Things-miljöer. Featureless- modeller förlitar sig på de starka nätverksstrukturella mönstren för botnet-attacker, medan heterogena modeller avser att fånga mer komplexa mönster genom att differentiera flera typer av noder och kanter i grafer. Potentialen hos Temporal Graph Neural Network undersöks också på dynamiska grafer. Jämförelsen av resultaten av de studerade egenskaperna gör det möjligt att diskutera potentialen hos implementeringar av grafneurala nätverk för intrångsdetektering i samband med dataset med hög obalans. Denna avhandling undersöker potentialen hos olika grafstrukturer i Graph Neural Network-modeller för att förbättra prestanda för upptäckt av botnet-attacker, vilket ger både teoretiska framsteg och praktiska lösningar för att säkra IoT-nätverk.

Place, publisher, year, edition, pages
2024. , p. 109
Series
TRITA-EECS-EX ; 2024:806
Keywords [en]
Cybersecurity, Graphs, Graph neural networks, Intrusion detection, Botnet, Internet of things, GCN, GAT, TGNN
National Category
Computer and Information Sciences
Identifiers
URN: urn:nbn:se:kth:diva-360311OAI: oai:DiVA.org:kth-360311DiVA, id: diva2:1939963
Supervisors
Examiners
Available from: 2025-02-27 Created: 2025-02-25 Last updated: 2025-02-27Bibliographically approved

Open Access in DiVA

fulltext(2198 kB)69 downloads
File information
File name FULLTEXT02.pdfFile size 2198 kBChecksum SHA-512
b034cc47a766c39a54cb845a70373c7dddbe9a14236766ec6943e7dacbd0a2c7f3808ef677ae0dd99083faa806f7a3086e4f5ea29a29b93d3ccc2fdecc306e04
Type fulltextMimetype application/pdf

By organisation
School of Electrical Engineering and Computer Science (EECS)
Computer and Information Sciences

Search outside of DiVA

GoogleGoogle Scholar
Total: 69 downloads
The number of downloads is the sum of all downloads of full texts. It may include eg previous versions that are now no longer available

urn-nbn

Altmetric score

urn-nbn
Total: 436 hits
CiteExportLink to record
Permanent link

Direct link
Cite
Citation style
  • apa
  • ieee
  • modern-language-association-8th-edition
  • vancouver
  • Other style
More styles
Language
  • de-DE
  • en-GB
  • en-US
  • fi-FI
  • nn-NO
  • nn-NB
  • sv-SE
  • Other locale
More languages
Output format
  • html
  • text
  • asciidoc
  • rtf